Auto-rigging places a skeleton inside a mesh and binds the surface to it, so the model can be animated. It works reliably on exactly one category of model, and fails predictably outside it.
The requirement: a humanoid in a T-pose
- Humanoid. One head, two arms, two legs, an upright torso. The rigger is matching your mesh against a known skeleton layout; if the shape does not correspond, there is nothing to match.
- T-pose. Arms straight out to the sides, roughly horizontal. Legs straight, slightly apart. Facing forward.
An A-pose — arms angled down toward the hips — is the most common failure, and the most surprising one, because the model looks perfectly riggable. The joints land in the wrong places and the shoulders deform badly.
What is not riggable this way
Quadrupeds, vehicles, props, weapons, buildings, and anything abstract. Not because they cannot be animated, but because they have no standard skeleton to match — a horse and a spider and a crane all need bespoke rigs.
Getting a model into a T-pose
If you are generating the character, ask for the pose in the prompt or supply reference images in T-pose. It is far easier to generate a T-pose than to re-pose a finished mesh.
If you already have a posed model, re-posing it means rigging it first — which is the problem you were trying to solve. Usually it is faster to regenerate.
After rigging
A skeleton is not an animation. You get bones and skin weights; motion comes from animation clips, either authored or retargeted from a library. Retargeting works when the skeletons share a layout, which is one of the reasons a standard humanoid rig is worth having.
Expect to inspect the weights. Automatic skinning is good at the torso and progressively worse toward the extremities — fingers, jaw and anything with thin overlapping geometry usually need a pass by hand.